Файл: [D._Kim_Rossmo]_Geographic_Profiling(BookFi.org).pdf

© 2000 by CRC Press LLC

9.2.1 Methodology

9.2.1.1

Serial Killer Data

The first database (henceforth referred to as the FBI serial killer data set)
contains macrolevel information on serial killers. The main source for this
data was an FBI NCAVC 

Lexus/Nexus

 newspaper computerized topical search

file on serial murderers in the U.S. The crimes are therefore primarily, though
not exclusively, North American. An attempt was made to clean and verify
the information in the original list, and to eliminate those cases that are closer
to a spree or mass murder classification. To this end, the FBI definition of
serial murder was used (see Ressler et al., 1988). While there was no attempt
to be comprehensive, an effort was made to update the data and include well-
known foreign serial killers. The final list consists of 225 cases, including
most well-known serial killers. Information in the FBI serial killer data set
includes: (1) offender name; (2) moniker or media name; (3) sex; (4) asso-
ciates; (5) number of confirmed victims; (6) number of suspected victims;
(7) start date of murders; (8) end date of murders; and (9) cities, states or
provinces, and countries of operation.

The data set on serial killers served two analytic purposes. First, it was

the means by which frequencies, averages, and rates were computed for the
macrolevel variables. Regional variation in serial murder was also examined
through state comparisons of counts and rates (adjusted for both population
and overall murder). Second, this list was the source for the microlevel serial
murder data set. One of the critical elements of the selection criteria involved
choosing those cases where the serial murderer’s activity space and hunting
style produced target patterns that are amenable to geographic profiling. This
ultimately led to the development of the predator hunting typology, a nec-
essary step in the articulation of the applicability and limits of geographic
profiling. The larger data set made this process feasible.

9.2.1.2

Newspaper Sources

While the use of newspaper data in serial murder research is common, a few
cautions are worth noting. Egger (1990) states that “it would appear that the
mass media are currently the only source from which to quantify serial
murders in this country [U.S.]” (p. 11). He searched 

The New York Times

Index from January 1978 to June 1983 for cases of serial killers (1984, 1990).
Simonetti (1984) perused the same index and associated microfilm files for
stories of serial murder cases from January 1970 to November 1983. Jenkins
(1988a, 1989) researched newspaper archives, including 

The New York Times

Index, to compile a list of major cases of serial murder in the U.S. from 1900
to the present, finding just as much journalistic interest in this topic in the
early part of the century as in the later.

© 2000 by CRC Press LLC

Levin and Fox (1985) searched indices of six major American newspapers

for serial and mass murder cases involving a minimum of four victims from
1974 to 1979. Hickey (1997) identified serial murder cases from 1800 to 1995
through a variety of sources including newspapers, journals, biographies,
interviews, bibliographies, and computer abstract searches. Anytime the
media are used as a source of information, it becomes important to ascertain
the existence of possible reporting biases. Kiger (1990) warns that “newspaper
stories are problematic sources of data because they are obviously dependent
upon editorial decisions and because they may sensationalize (not to mention
glorify) this phenomenon in order to increase circulation” (p. 37). The media
may be inclined to focus on particularly heinous cases or those involving a
large number of victims, and ignore African-American (Jenkins, 1993a) and
foreign offenders. In his analysis of the Henry Lee Lucas serial murder case,
Egger (1990) found newspaper accounts were often inaccurate in their
descriptions of details and had to be treated as suspect.

Kiger (1990) further cautions that a string of murders must first be iden-

tified as part of the same series before the existence of a serial killer will even
be suspected (see the earlier discussion on linkage blindness). Newspaper
reporting inaccuracies and definitional problems further complicate media
source research. For example, upon perusing case details in microfilmed news-
paper articles, Simonetti (1984) found some so-called incidents of serial mur-
der were actually mass murders. Jenkins (1989) discusses the problems of
historical changes in media practice and coverage, and the tendency towards
greater coverage of events that occur in large urban areas. He notes major
metropolitan papers expanded their coverage of regional issues in the 1950s,
and since 1960 the proportion of national newspaper chains has tripled
(1992b). In subsequent research, however, he concludes that such concerns
do not present major research difficulties or significant data biases (1989).

These problems suggest a potential research bias. Crimes of repetitive

killers who hunt over wide geographic regions may, because of linkage
blindness, be unidentified as serial murders. Those attacks lacking a consis-
tent M.O. and spread out over long periods of time are even less likely to
be connected. Some serial murders may be officially known only as a number
of unrelated missing persons reports. Such crimes fall into the “dark figure”
of serial murder and are not open to traditional methods of information
gathering, possibly skewing data collection towards more geographically
confined serial murderers. But this begs the question; if a murder series is
not identified in the first instance, then there is no need to assist in its
investigation. It is also possible that purported but unidentified serial mur-
derers, such as Jack the Ripper or the Green River Killer, never actually
existed. While that conclusion is unlikely in these particular cases, it is
certainly feasible the press or police might incorrectly link a number of

© 2000 by CRC Press LLC

unconnected homicides (see Jenkins, 1989). This is apparently what hap-
pened with a 1984 to 1985 series of murders of women in Fort Worth, Texas
(Jenkins, 1993b).

9.2.1.3

Offender, Victim, and Location Data

The second database (henceforth referred to as the SFU serial murder data
set) contains microlevel information on offenders, victims, and locations for
selected serial murder cases. It is based on a sample drawn from the FBI serial
killer list, which served as a reasonable approximation of the overall popu-
lation. While an accurate accounting of all serial murderers does not exist
and most likely never will, the best efforts to date suggest the size of the
modified FBI list is on the right order of magnitude (see Cavanagh, 1993;
Hickey, 1997).

Specific criteria were employed to structure the sampling procedure.

Only serial killers responsible for five or more victims, after 1960, who were
residentially stable, operated by themselves or with a single partner, and who
hunted in certain ways, were included (see Rossmo, 1995a, for complete
details of the sampling methodology). The selection process helped to define
the appropriate limits of the analysis and prevent misapplication. To have
value in the investigation of an unsolved serial murder case, such a classifi-
cation must be determinable from information likely to be known to police
prior to the killer’s apprehension. Research parameters were developed with
this consideration in mind.

A sample of 10 cases was randomly drawn from those that met the

selection criteria and used to examine the microlevel geography of serial
murder. An additional three cases that violated various selection criteria were
also studied in order to assess the impact of these violations. The resulting
13 cases served as the initial test group for the criminal geographic targeting
algorithm used in geographic profiling.

Information in the SFU serial murder data set includes: (1) date of

murder; (2) day of murder; (3) time of murder; (4) sequential case number;
(5) weapon; (6) vehicle use; (7) degree of offender organization; (8) serial
killer classification; (9) hunting typology; (10) victim traits (specific or non-
specific); (11) victim selection (random/nonpatterned or nonrandom/pat-
terned); (12) victim activity; (13) offender residence (address, and spatial
coordinates); (14) offender workplace (address, and spatial coordinates); and,
for all known crime locations connected to each murder, (15) area descrip-
tion; (16) neighbourhood type; and (17) spatial coordinates. Appendix B
contains the serial murderer, victim, and crime location data coding forms.
Data coding procedures and guidelines can be found in Rossmo (1995a).

Through the use of a relational database management system, crime

locations were hierarchically connected through the common murder, and

© 2000 by CRC Press LLC

murders linked through the common killer. Further data items were calcu-
lated from these relationships, such as the distance from residence to crime
site. A total of 13 serial murder cases, comprising 15 serial killers, 178 victims,
and 347 crime locations constitute the SFU data set.

9.2.2 Serial Killer Characteristics

A breakdown of the characteristics for the 225 serial killers in the FBI data
set is given in 

Table 9.2

. The complete list is presented in Appendix A,

Table A.1

.

As there were 14 cases that remained unsolved with the identity of the

offender unknown, the proportion of unsolved cases is 7.3% (there are 193
unique cases in the data set). The percentage of female offenders is compa-
rable to that found in previous research (cf. Hickey, 1997; Newton; 1992), as
is the estimate for mean duration of murder activity (cf. Hickey, 1991; Jen-
kins, 1988b). One-quarter of the serial murderers operated with a partner,
and the proportion of cases with more than one offender is 11.9%. This
estimate is at the low end of the range found in previous studies (cf. Hickey,
1997; Jenkins, 1990; Simonetti, 1984; Newton, 1992). Team killers averaged
1.7 partners.

50

Figure 9.1 

presents the distributions for number of confirmed and sus-

pected victims by case. The mean for the suspected number of victims was
calculated after eliminating those cases that claimed 100 or more victims, as

Table 9.2 Serial Killer Characteristics

Characteristic

Results

Offender Identity

Known

93.8% (211)

Unknown

6.2% (14)

Sex

Male

90.7% (204)

Female

9.3% (21)

Co-Killer

Operated Alone

75.6% (170)

Operated With Partner

24.4% (55)

Mean Duration of Murder Activity

4.4 Years

Mean Number of Confirmed Victims

9.7

Mean Number of Suspected Victims

13.3

Mean Number of Different Cities

2.8

Mean Number of Different States

1.7

50 

Interestingly, the hunting efficiency of solitary lions ranges from 8 to 19%, but increases

to 30% in cooperative hunts (Barnard, 1984a; Schaller, 1972).

© 2000 by CRC Press LLC

these figures were not considered reliable (the mean is 18.7 otherwise). The
estimate for the mean number of different cities should be regarded with
caution as the data appear to contain listings for both metropolitan areas
(e.g., New York) and individual cities (e.g., Brooklyn, Queens).

9.2.2.1

State Comparisons

Many researchers have noted regional variations in multiple murder, but
there is little consensus as to where the high rate areas are. The Pacific
Northwest is often held out to be the location with the most incidents of
serial murder (Egger, 1990; Mathers, 1989). Cavanagh (1993) found that
39.6% of serial murder victims were from the Pacific subregion, a proportion
more than twice as great as the next highest subregion. Levin and Fox (1985)
observe that multiple murderers usually strike in urban areas, most likely in
New York, Texas, or on the West Coast, particularly Southern California.
They are least likely to attack in the Deep South, with the exception of Texas.
But 64% of the female serial killers in Keeney and Heide’s (1994c) study were
from the South, most commonly Florida (29%).

Hickey (1997) found a lack of regionality in serial murder, though Cal-

ifornia reported over half again the number of serial murder cases as New
York, the next highest state. In Cavanagh’s (1993) analysis, California had
over four times the number of victims of New York, the next highest state.
Hickey observes that population density, particularly in metropolitan areas,
appears to be a more important correlate than region. Jenkins (1990) suggests
that the geographical concentration of serial murder in the western U.S. may
be partially attributable to “a culture of casual predatory sexuality,” and
differential access to vice facilities in Californian cities. Levin and Fox (1985)

Figure 9.1

Confirmed and suspected victim numbers by case.